Room9

CKAD - Application observability and maintenance (probes and health checks) 본문

Kubernetes

CKAD - Application observability and maintenance (probes and health checks)

Room9_ 2022. 2. 15. 23:21

Probe

Probe는 Kubelet을 통하여 설정된 주기에 의해 컨테이너를 진단하는 역할을 담당한다. 

  • Readiness Probe
  • Liveness Probe
  • Startup Probe

Readiness Probe

Readiness Probe를 사용하여 컨테이너가 언제 준비되는지 진단할 수 있다.

Pod는 내부의 모든 컨테이너가 Ready 상태가 되면 트래픽을 받기 시작한다. Readiness Probe는 컨테이너가 트래픽을 처리할 준비가 되었는지 진단을 하고 Success 상태가 되면 서비스와 연결하여 트래픽을 처리하게 된다.

Readiness probe가 실패한다면, 엔드포인트 컨트롤러는 파드에 연관된 모든 서비스들의 엔드포인트에서 파드의 IP 주소를 제거하여 트래픽 전달을 하지 않는다.

Liveness Probe

Liveness Probe는 컨테이너를 언제 다시 시작해야하는지 진단한다. Liveness Probe에서 진단이 실패하게 되면 Kubelet은 해당 컨테이너를 죽이고, 컨테이너는 재시작 정책의 대상으로 변경되고, 재시작된다. 서비스의 가용성을 높일 수 있다.

Startup Probe

컨테이너 내의 어플리케이션이 시작되었는지를 진단한다. Startup Probe가 설정된 경우에는, 진단이 성공할 때 까지 컨

다른 Probe는 활성화 되지 않는다. Startup Probe가 실패하면, kubelet은 컨테이너를 죽이고, 컨테이너는 재시작 정책에 따라서 처리 된다.

KodeKloud

Q1. We have deployed a simple web application. Inspect the PODs and Services

root@controlplane:~# kubectl get pods
NAME              READY   STATUS    RESTARTS   AGE
simple-webapp-1   1/1     Running   0          3m33s

Q2. View the application by clicking on the 'Web Portal' link above your terminal

Q3. A test script is provided that sends multiple requests to access the web application.

root@controlplane:~# ll
total 44
drwx------ 1 root root 4096 Feb 15 13:54 ./
drwxr-xr-x 1 root root 4096 Feb 15 13:52 ../
-rw-r--r-- 1 root root 3236 Feb 15 13:54 .bashrc
drwx------ 2 root root 4096 Feb 15 13:54 .cache/
drwxr-xr-x 3 root root 4096 Feb 15 13:54 .kube/
-rw-r--r-- 1 root root  148 Aug 17  2015 .profile
drwx------ 2 root root 4096 Feb 15 13:54 .ssh/
-rwxr-xr-x 1 root root  114 Feb 12 02:55 crash-app.sh*
-rwxr-xr-x 1 root root  216 Feb 12 02:55 curl-test.sh*
-rwxr-xr-x 1 root root  123 Feb 12 02:55 freeze-app.sh*

Q4. 

root@controlplane:~# kubectl get pods
NAME              READY   STATUS    RESTARTS   AGE
simple-webapp-1   1/1     Running   0          5m14s
simple-webapp-2   1/1     Running   0          20s

Q5. Notice the errors in the output of the script.

root@controlplane:~# ./curl-test.sh 
Failed

Failed

Failed

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Q6. Update the newly created pod 'simple-webapp-2' with a readinessProbe using the given spec

Spec is given on the below. Do not modify any other properties of the pod.

  • Pod Name: simple-webapp-2
  • Image Name: kodekloud/webapp-delayed-start
  • Readiness Probe: httpGet
  • Http Probe: /ready
  • Http Port: 8080
root@controlplane:~# kubectl get pods simple-webapp-2 -o yaml > 2.yaml
apiVersion: v1
kind: Pod
metadata:
  labels:
    name: simple-webapp
  name: simple-webapp-2
  namespace: default
spec:
  containers:
  - env:
    - name: APP_START_DELAY
      value: "80"
    image: kodekloud/webapp-delayed-start
    imagePullPolicy: Always
    name: simple-webapp
    ports:
    - containerPort: 8080
      protocol: TCP
    readinessProbe:
      httpGet:
        path: /ready
        port: 8080
root@controlplane:~# kubectl delete pod simple-webapp-2
root@controlplane:~# kubectl apply -f 2.yaml 
pod/simple-webapp-2 created

Q7. Run the script again to test the web application. Notice that none of the requests fail now. Though all the requests hit the same POD.

root@controlplane:~# ./curl-test.sh 
Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Q8. Wait for the new POD to be ready and run the test again to see traffic distributed between both the PODs.

root@controlplane:~# ./curl-test.sh 
Message from simple-webapp-2 : I am ready! OK

Message from simple-webapp-2 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-2 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Q9. What would happen if the application inside container on one of the PODs crashes?

Try it by accessing url /crash of the application in your browser or run the crash-app.sh script. Then check the status of POD. Run the curl-test to see if users are impacted.

root@controlplane:~# kubectl get pods
NAME              READY   STATUS    RESTARTS   AGE
simple-webapp-1   1/1     Running   0          15m
simple-webapp-2   1/1     Running   0          2m34s
root@controlplane:~# ./crash-app.sh 
Message from simple-webapp-1 : Mayday! Mayday! Going to crash!
root@controlplane:~# kubectl get pods
NAME              READY   STATUS    RESTARTS   AGE
simple-webapp-1   1/1     Running   1          15m
simple-webapp-2   1/1     Running   0          2m43s

> The container inside the POD is restarted

Q10. When the application crashes, the container is restarted. During this period the service directs users to the available POD, since the POD status is not READY.

Q11. What would happen if the application inside container on one of the PODs freezes?

Try it by accessing url /freeze of the application in your browser or run the freeze-app.sh script. Then check the status of POD. Run the curl-test to see if users are impacted.

 > New users are impacted

Q12. Update both the pods with a livenessProbe using the given spec

Delete and recreate the PODs.

  • Pod Name: simple-webapp-1
  • Image Name: kodekloud/webapp-delayed-start
  • Liveness Probe: httpGet
  • Http Probe: /live
  • Http Port: 8080
  • Period Seconds: 1
  • Initial Delay: 80
  • Pod Name: simple-webapp-2
  • Image Name: kodekloud/webapp-delayed-start
  • Liveness Probe: httpGet
  • Http Probe: /live
  • Http Port: 8080
  • Initial Delay: 80
  • Period Seconds: 1
apiVersion: v1
kind: Pod
metadata:
  labels:
    name: simple-webapp
  name: simple-webapp-1
  namespace: default
spec:
  containers:
  - image: kodekloud/webapp-delayed-start
    imagePullPolicy: Always
    name: simple-webapp
    ports:
    - containerPort: 8080
      protocol: TCP
    livenessProbe:
      httpGet:
        path: /live
        port: 8080
      initialDelaySeconds: 80
      periodSeconds: 1
apiVersion: v1
kind: Pod
metadata:
  labels:
    name: simple-webapp
  name: simple-webapp-2
  namespace: default
spec:
  containers:
  - env:
    - name: APP_START_DELAY
      value: "80"
    image: kodekloud/webapp-delayed-start
    imagePullPolicy: Always
    name: simple-webapp
    ports:
    - containerPort: 8080
      protocol: TCP
    readinessProbe:
      httpGet:
        path: /ready
        port: 8080
    livenessProbe:
      httpGet:
        path: /live
        port: 8080
      initialDelaySeconds: 80
      periodSeconds: 1
root@controlplane:~# kubectl apply -f 1.yaml -f 2.yaml 
pod/simple-webapp-1 created
pod/simple-webapp-2 created
root@controlplane:~# kubectl get pods
NAME              READY   STATUS              RESTARTS   AGE
simple-webapp-1   0/1     ContainerCreating   0          3s
simple-webapp-2   0/1     ContainerCreating   0          3s

Q12. Freeze the application again and notice what happens. When a POD freezes, it would be restarted automatically.

root@controlplane:~# ./curl-test.sh 
Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

Message from simple-webapp-1 : I am ready! OK

root@controlplane:~# kubectl get pods
NAME              READY   STATUS    RESTARTS   AGE
simple-webapp-1   1/1     Running   0          2m37s
simple-webapp-2   0/1     Running   1          2m37s

 

'Kubernetes' Related Articles more
Comments